Abstra t- The aim of this paper is to show that the ommon belief, in the evolutionary ommunity, that evaluation time usually takes over 90 % of the total time, is far from being always true. In fa t, many real-world appli ations showed a mu h lower per entage. This raises several questions, one of them being the balan e between tness and operators om-putational omplexity: what is the use of elaborating smart evolutionary operators to redu e the number of evaluations if as a result, the total omputation time is in reased? I. Introdu tion In a re ent resear h work, we ompared two evolu-tionary libraries with the help of testben h fun tions[20℄. During the extensive tests that were involved, we ame a ross the fa t that the time spent by an evolutionary algorithm in the evaluation / tness fun tion was not as high as expe ted. In fa t, it appeared that the overhead indu ed by ge-neti operators (sele tion, mutation, re ombination, et.) was far from being negligible in many ases. Therefore, it ame down to us that people may not always be opti-mising the right pie e of ode. After shortly dis ussing EA ode optimisation, we will des ribe four very dierent evolutionary algorithms where the evaluation fun tion takes from 98 % of the total time down to 26 % only. We will then use ommon testben hes to try to show that in some ases, over-optimising evolutionary opera-tors may not be a ni e idea after all. II. Optimising the EA ode Many people have been working on the optimisation of the ode of Evolutionary Algorithms, with the result that many papers are available on the subje t. As s ientists, resear hers on entrated on the optimisation of the algo-rithm itself using the widespread knowledge on this eld. As an example, many resear h works were ondu ted on optimising the size of the population [15℄, [2℄: Fu hs[11℄ 1 European Commission IST Programme 1999-12679 (Future and

Abstract not found

à la Faculté des études supérieures de l’Université Laval dans le cadre du programme de doctorat en informatique pour l’obtention du grade de Philosophiae doctor (Ph.D.)

This paper presents a technique to select a representative set of test cases from a test suite that provides the same coverage as the entu-e test suite This selectlon m performed by ldentlfymg, and then ehmmating, the redundant and obsolete test cases in the test suite. The representatl ~ e set replaces the orgznal test mute and thus, potential y produces a smaller test suite The representative set can also be used to identify those test cases that should be rerun to test the program after it has been changed. Our techmque us independent of the testing methodology and only requu-es an association between a testing requirement and the test cases tb at satisfy the requirement We i] lustrate the technique using the data flow testing methodology. The reduction that is possible with our technique is illustrated by experimental results,

C'est dans le bureau d'Elisabeth Szuyska, de réparties mi-plaisantes mi-sérieuses, que le nom de CONDOR survint. Elisabeth fut ainsi associée à la naissance même du séminaire, auquel elle participa activement à ses débuts. Elle préparait chaque année l'édition des Actes, rappelant à l'ordre les auteurs quand il le fallait, rassemblant les textes et les corrigeant avec le soin et la minutie qu'elle mettait en toute chose. Elle nous a quittés en 1996 et sa présence discrète et efficace nous manque aujourd'hui.. Ces Actes, avec les imperfections qu'ils peuvent comporter, lui sont dédiés.

Written by an international assembly of distinguished philosophers, the Blackwell Philosophy Guides create a groundbreaking student resource – a complete critical survey of the central themes and issues of philosophy today. Focusing and advanc-ing key arguments throughout, each essay incorporates essential background material serving to clarify the history and logic of the relevant topic. Accordingly, these volumes will be a valuable resource for a broad range of students and readers, including professional philosophers.

From 1960-1983, the standard deviation of annual growth rates in real GDP in the United States was 2.7%. From 1984-2001, the corresponding standard deviation was 1.6%. This paper investigates this large drop in the cyclical volatility OF real economic.activity. The paper has two objectives. The first is to provide a comprehensive characterization of the decline in volatility using a large number of U.S. economic time series and a variety of methods designed to describe time-varying time series processes. In so doing, the paper reviews the literature on the moderation and attempts to resolve some of its disagreements and discrepancies. The second objective is to provide new evidence on the quantitative importance of various explanations for this "great moderation". Taken together, we estimate that the moderation in volatility is attributable to a combination of improved policy (20-30%), identifiable good luck in the form of productivity and commodity price shocks (20-30%), and other unknown forms of good luck that manifest themselves as smaller reduced-form forecast errors (40-60%).

Thèse de doctorat présentée en vue de l'obtention du grade de Docteur de l'Université Pierre et Marie Curie Spécialité Physique de la particule à la matière condensée par

Lexical-semantic resources, including thesauri and WORDNET, have been successfully incorporated into a wide range of applications in Natural Language Processing. However they are very difficult and expensive to create and maintain, and their usefulness has been severely hampered by their limited coverage, bias and inconsistency. Automated and semi-automated methods for developing such resources are therefore crucial for further resource development and improved application performance.

Méccanismmes de mise en place du ma atériel Tunnisie sepptentri ionale: exem mple du